A question about time dimension for netcdf files

[Don Hooper]

Haibin Li,

Yes and no.  Sure, it will check the "calendar" attribute on the data variable
for the value "365_day_year" or "noleap" and put itself into a 365 day
perpetual calendar mode.  Alas, however, it still uses the Udunits library
(which does not, as far as I can make out, support such a calendar) to
interpret the first two time values in the time coordinate variable to
orient the time axis.  This leads to problems as you might imagine.  The
workaround is to write a minimal DDF (Data Descriptor File) for the
xdfopen interface consisting of just DSET and TDEF lines.  If you put the
full time coordinate information into the TDEF entry, things should work.
Just to leave less to chance, it's a good idea to add a line between DSET
and TDEF that says "OPTIONS 365_day_calendar".  That should work.  This
issue doesn't arise (I don't believe) in version 1.9's "dtype netcdf"
interface, as you would be writing a full DDF with a full TDEF line
for that interface, instead of using much in the way of the metadata
from the NetCDF file.

Good luck,
-Hoop
> From owner-gradsusr at LIST.CINECA.IT  Wed Jan 19 20:01:55 2005
> Reply-To: Haibin Li <haibin1977 at gmail.com>
>
> Hello,
> Recently I am analyzing some model outputs which are netcdf format. It
> seems some files are recognized successfully and others are not in
> terms of the time step. Can someone tell me whether GrADS support the
> netcdf files with "no-leap" year convention?  Thanks
>
>
> Yours,
> Haibin
>